Daily time series of common single word searches in AltaVista and NorthernLight

Abstract

Over the past year the AltaVista search engine provided irregular numbers of
\nhits. This is shown by collecting a time series of daily searches. The new
\nAltaVista (on the Web since October 25, 1999) is much more stable, but not
\nall problems have been eliminated. The same searches performed on
\nNorthernLight gave stable results. It is suggested to use filtering techniques,
\nsuch as median filtering, when dealing with irregular time series. An attempt
\nis made to estimate the growth of the Web (as covered by a search engine),
\nbased on neutral single word searches. It turned out that this method
\nsystematically underestimates the growth of the Web. Collecting time series
\nshould be an essential part of Internet research